Chemical Synthesis Reveals Pathogenic Role of N-Glycosylation in Microtubule-Associated Protein Tau

Abstract

Alzheimer’s disease (AD) is a neurodegenerative disorder characterized by the accumulation of tau protein aggregates. In this study, we investigated the effects of N-glycosylation on tau, focusing on its impact on aggregation and phase behavior. We chemically prepared homogeneous glycoproteins with high-mannose glycans or a single N-acetylglucosamine at the confirmed glycosylation sites in K18 and 2N4R tau. Our findings reveal that N-glycosylation significantly alters biophysical properties and potentially cellular functions of tau. Small glycans promote tau aggregation and liquid–liquid phase separation (LLPS), while larger glycans reduce these effects. High mannose glycans at N410 enhance phosphorylation by GSK3β, suggesting a pathological role in AD. Functional assays demonstrate that N-glycosylation does not impact microtubule polymerization dynamics but modulates aggregation kinetics and morphology. This research underscores the importance of glycosylation in tau pathology and opens new avenues for therapeutic interventions targeting glycan processing.

Graphical Abstract

INTRODUCTION

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder and a major cause of dementia.¹ Mounting evidence suggests that the disease develops decades before clinical symptoms appear, emphasizing the importance of studying the early changes. Although not fully understood, amyloid plaques and intraneuronal deposits of neurofibrillary tangles composed of paired helical filaments (PHFs) of aggregated tau are the two pathological hallmarks of the disease.² Microtubule-associated protein tau is an intrinsically disordered protein, and various posttranslational modifications (PTMs) regulate its function.^3,4 While hyperphosphorylation is closely associated with tau self-aggregation,^5–7 PHF formation has not been demonstrated by hyperphosphorylation alone, pointing to the interplay of phosphorylation with other PTMs as drivers of fibrilization. Hyperphosphorylated tau and PHF-tau were found to be N-glycosylated in AD but not in healthy brains.^8,9 Three potential glycosylation sites were predicted in the longest 2N4R isoform of tau (N167, N359, and N410), but only two (N359 and N410) or one (N410) have been confirmed in neuroblastoma cells and AD brains, respectively (Figure 1A).^9–12 Prior studies have also established that soluble and aggregated forms of tau have different glycan structures, with high mannose, truncated, fucosylated, and biantennary bisecting glycans present.¹³ Site-mutagenesis studies demonstrated aggravated AD-like neurodegenerative phenotypes in an insect model, further validating the pathological functions of N-glycosylation.¹¹ The observation that healthy and AD tau samples have distinct glycan compositions suggests their unique physiological functions.¹⁴ However, the molecular underpinnings underlying of the role of N-glycosylation in tau biology remain unknown. Here, we show that N-linked glycans impact the aggregation and phase behavior of tau glycoproteins. By combining synthetic chemistry with biophysical and structural tools, we establish that N-linked glycans are drivers of tau pathology. Furthermore, by producing a library of tau glycoforms, we show that the size of the glycan follows a general trend where small glycans promote aggregation, liquid–liquid phase separation (LLPS), and seeding, and these effects diminish or even reverse upon increasing the oligosaccharide size. However, the effects at the glycosylation sites tend to vary with the high mannose glycans located outside of the AD ordered core at N410 promoting phosphorylation by GSK3β at residues in proximity. Taken together, these results support the notion that glycosylation outside of the ordered core may promote AD tau pathology and explain the differences in the glycan composition between soluble and aggregated tau in AD brains. Our findings bridge an important knowledge gap in pathobiology of tau and offer a molecular insight into potential therapeutic strategies based on perturbation of glycan processing.^15,16

Figure 1.

(A) Domain organization of 2N4R Tau, location of N-linked glycans, and the structures of the glycans from AD patients. (B) Schematic of the N-glycosylated Tau constructs used in this study.

RESULTS

To investigate the effects of N-glycosylation on tau, we used chemical methods to prepare homogeneous glycoproteins with a high-mannose glycan or a single N-acetylglucosamine (GlcNAc), both installed at the confirmed glycosylation sites in K18 and 2N4R tau (Figure 2). To this end, we first developed a total synthesis of Man₅GlcNAc₂, the most predominant glycan found in tau from AD brains (Figure 1B).¹³ Our synthesis features trisaccharide 2.1¹⁷ that was subjected to α-manosylation with glycosyl acceptor 2.2 and the regioselective acetal reduction with BH₃/Bu₂BOTf producing tetrasaccharide 2.3. A convergent attachment of trisaccharide 2.4 was carried out next in 50% yield followed by a four-step global deprotection sequence (deacetylation with NaOMe, phthalimide removal, N-acetylation with Ac₂O, and debenzylation with H₂/Pd(OH)₂) furnishing the unprotected Man₅GlcNAc₂. This oligosaccharide was then converted into anomeric amine 2.6 installed with β-stereoselectivity using NH₄HCO₃.¹⁸

Figure 2.

Preparation of the glycans and N-linked glycopeptides. (A) Chemical synthesis of Man₅GlcNAc₂ amine. Reagents and conditions: a. NIS, AgOTf, 72%; b. Bu₂BOTf, BH₃·THF, 59%; c. NIS, AgOTf, 68%; d. NaOMe, MeOH; e. ethylenediamine then Ac₂O, 48%; f. H₂, Pd(OH)₂ then NH₄HCO₃, H₂O, 87%; (B) Synthesis of key glycopeptide fragments. Glycopeptides prepared with solid phase glycopeptide synthesis with 2.8: automated solid-phase peptide synthesis then 2% (v/v) hydrazine hydrate, DMF, 23 °C, 2 h; then TFA:H₂O:thioanisole:TIPSH, 23 °C, 3 h. Reagents and conditions: g. glycosyl amine 2.6 or 2.7, protected peptide, HATU, DIPEA, DMF, 23 °C, 6 h; then TFA:H₂O:thioanisole:TIPSH, 23 °C, 3 h.

With the glycosyl amine in hand, we set out to prepare the key glycopeptide fragments (Figure 2B). First, we investigated solid phase glycopeptide synthesis and introduced N-acetylglucosamine into peptides Tau(322–372) 2.17, Tau(358–389) 2.12, and Tau(390–441) 2.18 with Fmoc-Asn(GlcNAc)–OH 2.8 as the key building blocks. However, when the same approach was attempted to install Fmoc-Asn(Man₅GlcNAc₂)–OH onto the Tau(322–372) fragment, no coupling was observed due to the large size of the glycan. The Lansbury aspartylation overcomes these limitations by engaging an anomeric amine in a direct reaction with a protected oligopeptide bearing a free side chain of aspartic acid.¹⁹ The efficiency of this transformation can be improved if the oligopeptide contains a pseudoproline at the NxS/T position, which disfavors problematic aspartimide cyclization.^20,21 Unfortunately, we found that the installation of 2.6 on the 51-amino acid long segment of Tau(322–372) was challenging most likely due to the length limitations characteristic of the Lansbury reaction. Therefore, we prepared a shorter oligopeptide Tau(358–372) 2.11 for the synthesis of K18 and efficiently merged it with chitobiose (2.7) and Man₅GlcNAc₂ (2.6) amines in 85% and 33% yield, respectively. Other segments for the synthesis of 2N4R tau such as Tau(358–389) 2.10 and Tau(390–425) 2.9 were also aspartylated with 2.6 in synthetically acceptable yields (21–19%).

The completed syntheses of K18 glycoproteins involve three (for K18 and K18G) or four (for K18C and K18M) segments in which the two cysteine residues (C322 and C291) could be used for sequential C → N native chemical ligations (NCL)²² with the first reaction at the K321-C322 junction, followed by the K290-C291 position (Figure 3).^23–26 The additional L357-D358 junction was established with the diselenide-selenoester ligation (DSL).²⁷ The consideration that N-terminal Gln residues can self-cyclize prompted us also to use an N-terminal acetyl group, which prevents pyroglutamate formation and a C-terminal thioester for NCL in Tau(244-290) 3.6. The Tau(291-321) 3.5 with a N-terminal thiazolidine group and the corresponding thioester was obtained by a solution phase esterification with EtSH/PyBOP. The unmodified and GlcNAc Tau(322-372) 3.3 were synthesized on a Rink amide resin, albeit an additional step was needed to liberate the glycan hydroxyl functionalities. The Tau(322-357) 3.2 was used in the form of a selenoester installed with DPDS/P(n-Bu)₃.²⁸ Additionally, the Tau(358-372) segments 3.1 contain an N-terminal diselenide obtained by oxidative cleavage of a protected selenoether using TFA/DMSO,²⁷ a N-linked glycan at N359, and a C-terminal amide. All these units were then combined to assemble N-glycosylated K18 proteins with a sequence of reactions summarized in Figure 3A. First, the DSL between 3.1 and 3.2 followed by Acm removal using AgOAc afforded Tau(322-372) 3.3 in good yields.²⁹ These intermediates were next merged with Tau(291-321) 3.4 in MPAA buffer, and the Thz group removed with MeONH₂ furnishing Tau(291-372) 3.5.^30,31 Finally, ligation with Tau(244-290) 3.6 afforded K18 glycopeptides 3.7 (Figure 3B).

Figure 3.

Chemical synthesis of K18 glycoproteins. (A) Synthetic peptides used for the synthesis of K18 Tau and summary of the reactions yields after HPLC purifications. Reagents and conditions: DSL (step a): i. 6 M Gnd·HCl, 200 mM Na₂HPO₄, pH 6.3, 23 °C, 1 h; ii. extraction of DPDS with hexanes; iii. adjust to 2% (v/v) with hydrazine hydrate, pH 7.4, 23 °C, 10 min; iv. adjust to 125 mM TCEP·HCl, 12.5 mM DTT, pH 5.3, 23 °C, 10 min. Acm removal (step b): AgOAc (30 equiv), AcOH:H₂O (1:1), 23 °C, 6 h. NCL-Thz removal (step c): i. 200 mM MPAA, 6 M Gnd·HCl, 200 mM Na₂HPO₄, 20 mM TCEP·HCl, pH 7.0, 23 °C, 4 h; ii. adjust to 200 mM MeONH₂·HCl, pH 4.0, 23 °C, 4 h. NCL (step d): 200 mM MPAA, 6 M Gnd·HCl, 200 mM Na₂HPO₄, 20 mM TCEP·HCl, pH 7.0, 23 °C, 4 h. (B) HRMS (ESI) spectra of synthetic K18 glycoproteins.

The experience of working with K18 proteins was then applied to prepare full-length 2N4R tau glycoproteins using a semisynthetic approach from Tau (2-290) and Tau(291-441) (Figure 4A). In order to prepare synthetic Tau(291-441), two additional junctions (G389-A390C and L425-A426C) were utilized as the ligation sites^{26,27,32–36} producing peptides 4.1 and 4.4, which also facilitated the installation of the glycans.³⁶ In this approach, the unnatural thiol functionalities were removed through radical desulfurization to arrive at the native sequence.^37,38 Due to the size of the targeted proteins, expressed protein ligation (EPL)³⁹ was used to synthesize thioester Tau(2-290) 4.8, initially produced as a fusion protein with a C-terminal Mxe GyrA intein.⁴⁰

Figure 4.

Chemical synthesis of 2N4R Tau glycoproteins. (A) Synthetic peptides used for the synthesis of 2N4R Tau and summary of reactions yields after HPLC purifications. Reagents and conditions: DSL (step a): i. 6 M Gnd·HCl, 200 mM Na₂HPO₄, pH 6.3, 23 °C, 1 h; ii. extraction of DPDS with hexanes; iii. adjust the reaction mixture to 2% (v/v) hydrazine hydrate, pH 7.4, 23 °C, 10 min; iv. adjust the reaction mixture to 125 mM TCEP·HCl, 12.5 mM DTT, pH 5.3, 23 °C, 10 min. Thioesterification conditions (step b): i. NaNO₂ (10 equiv), 6 M Gnd·HCl, 100 mM NaH₂PO₄, pH 3.0, −15 °C, 25 min then ii. addition of 200 mM MESNa, 6 M Gnd·HCl, 200 mM Na₂HPO₄, pH 7.4, −15 °C to 23 °C, 2 h then iii. adjust to 50 mM TCEP·HCl. Native Chemical Ligation conditions (steps c and f): 200 mM MPAA, 6 M Gnd·HCl, 200 mM Na₂HPO₄, 20 mM TCEP·HCl, pH 7.0, 23 °C, 4 h. Desulfurization conditions (steps d and g): 100 mM GSH, 175 mM TCEP·HCl, 6 M Gnd·HCl, 200 mM HEPES, 20 mM VA-044, pH 7.0, 37 °C, 24 h. Acm removal conditions (steps e and h): AgOAc (30 equiv), AcOH:H₂O (1:1), 23 °C, 6 h. NCL-Thz removal (step (i): 200 mM MPAA, 6 M Gnd·HCl, 200 mM Na₂HPO₄, 20 mM TCEP·HCl, pH 7.0, 23 °C, 4 h then adjust the reaction mixture to 200 mM MeONH₂·HCl, pH 4.0, 23 °C, 4 h. Final NCL conditions (step j): 200 mM MPAA, 6 M Gnd·HCl, 200 mM Na₂HPO₄, 100 mM lysine, 20 mM TCEP·HCl, pH 7.0, 23 °C, 4 h. (B) HRMS (ESI) spectra of synthetic 2N4R Tau glycoproteins.

To assemble the glycoproteins, we prepared two segments, 4.2 and 4.5, in a series of ligations, desulfurizations, and protective group manipulations. Next, a reaction between these two segments in MPAA buffer proceeded smoothly, followed by desulfurization and Acm removal, affording Tau(322-441) 4.6. The final steps of this synthesis involved the union of 4.6 and Tau(291-321) 3.4, thiazolidine manipulations, and NCL with thioester 4.8. The glycoproteins were purified using sample displacement mode chromatography to remove heterogeneous impurities, and LC-MS traces indicate the proteins were obtained in high purities (Figure 4B).⁴¹

Following the synthesis, we characterized the solution behavior of tau proteins using a variety of biophysical and functional methods summarized in Figure 5. Protein N-glycosylation can promote local β-structures, and the glycans can assist the folding process.^42,43 Furthermore, molecular dynamics simulations suggest that N-glycosylated tau exists largely in a folded state with a higher propensity for β-structure than the unmodified protein.⁴⁴ The CD spectra of all tau constructs have a minimum at ~200 nm indicative of a disordered structure (Figure 5A). Small amounts of reversible secondary structures also occur as indicated by a negative patch at ~220 nm.⁴⁵ The CD spectra of WT and N359 tau are overlapping, analogous to similar observations for unmodified K18 and the glycoforms. Upon heating, WT 2N4R tau became less disordered. To test if N-glycosylation induces changes in folding, we plotted the θ ratios at 200/217 nm as a function of temperature. These values are similar for all proteins, and their linear appearance points to the equilibrium between an unfolded and folded state but irreversible hydrophobic folding is absent. The 200/217 ratios are ~1 unit higher for tau than K18 because tau has less secondary structure or is less hydrophobic than the aggregation prone repeat domain.

Figure 5.

Characterization of Tau glycoproteins. (A) CD spectra of N-glycosylated K18 and 2N4R Tau (0.4 mg/mL, 20 mM sodium phosphate, pH 7.4, 25 °C), temperature dependent CD spectra of 2N4R Tau, and the 200/217 nm θ ratio at various temperatures. (B) SDS-PAGE of K18 and 2N4R Tau glycoproteins (Coomassie brilliant blue). (C) Microtubule polymerization assay for K18 and 2N4R Tau. Assay conditions: 5 μM Tau(2-441) or 15 μM K18, 3 mg/mL tubulin, 80 mM PIPES, 2 mM MgCl₂, 0.5 mM EGTA, 1 mM GTP, pH 6.9, 37 °C (n = 3, the error bars represent the SD). (D) In vitro ERK2 and GSK3β phosphorylation of N-glycosylated 2N4R Tau. Conditions for ERK2:100 μM Tau, 1 μM ERK2, 2.5 mM ATP, 2 mM EGTA, 1 mM PMFS, 12.5 mM MgCl₂, 2 mM DTT, 50 mM NaCl, 50 mM HEPES (pH 7.4), 37 °C, 24 h. Conditions for GSK3β phosphorylation: 100 μM Tau, 1 μM GSK3β, 2 mM ATP, 5 mM EGTA, 1 mM PMFS, 5 mM MgCl₂, 2 mM DTT, 50 mM HEPES (pH 7.4), 30 °C, 24 h.

Normal tau from healthy brains exhibits greater mobility on SDS-PAGE compared to nonhyperphosphorylated tau from AD samples, a shift hypothesized to be due to N-linked glycans.⁴⁶ To test whether N-glycosylated tau has a reduced mobility compared to nonglycosylated tau, we characterized the proteins using SDS-PAGE (Figure 5B). SDS-PAGE analysis of K18 tau shows that each glycosylated construct has a different molecular weight. The visible shift in mobility correlates with the glycan size, as constructs with one, two, and seven sugar residues separate accordingly. The unmodified and monoglycosylated proteins have a similar shift on the gel. However, high mannose glycans result in higher apparent molecular weights than the monoglycosylated and unmodified tau forms.

Moving on to functional assays, we tested microtubule polymerization activity of tau glycoforms. Previous studies have shown that deglycosylation and subsequent dephosphorylation restore the microtubule polymerization activity of ex vivo AD p-tau and PHF tau.⁸ We therefore conducted microtubule assembly assays in the presence of N-glycosylated tau (Figure 5C). Microtubule polymerization was promoted to a similar extent in both unmodified and N-glycosylated variants, although 410G and 359M tau exhibited slightly reduced kinetics. For comparison, polymerization was most accelerated in the presence of unmodified K18, with the kinetics decreasing as the glycan size increased.

To investigate the potential interplay between N-glycosylation and phosphorylation, we mapped kinase-specific phosphorylation patterns of the full-length constructs (Figure 5D). We subjected 2N4R tau to in vitro phosphorylation with GSK3β and ERK2 and estimated the extent of phosphorylation at each site using mass spectrometry by quantifying the ratio of modified to unmodified peptide intensities. ERK2 phosphorylated tau with higher stoichiometry, while GSK3β had a lower stoichiometry (~5:1 and ~1:1 ratio of modification, respectively). ERK2 phosphorylation sites were located in the proline-rich region (PRR) and C-terminal domain (CTD), whereas GSK3β preferentially phosphorylated the PHF-1 epitope. Both kinases phosphorylated tau at identical residues across all five constructs, indicating that N-glycosylation does not alter the residue specificity. However, a high mannose glycan at N410 promoted GSK3β phosphorylation at S400 and S404, despite being located near the bulky saccharide. Previous studies comparing phosphorylation of ex vivo deglycosylated and dephosphorylated AD tau with healthy tau found that glycosylation promotes phosphorylation for cAMP kinase but reduces it for GSK3β and CDK5.^46,47 Our findings contrast with these prior results. We believe that this discrepancy is because the prior work used tau isolated from a primary patient sample. The dephosphorylation was likely incomplete, or the phosphatase has reduced activity for glycosylated material, therefore there are higher baseline phosphorylation levels.⁴⁸

Further studies were conducted to assess the aggregation properties of tau glycoproteins. In AD patients, the aggregated and soluble tau species both contain N-linked glycans; however, some of the PHF and soluble monomers remain nonglycosylated, raising questions about glycan involvement in aggregation and templated spreading.^8,46 Using fluorescent aggregation assays, we determined how the glycan type and modification sites affect the kinetics of amyloid formation (Figure 6).⁴⁹ For K18 glycoproteins, small glycans accelerate aggregation, and this trend reverses as the sugar size increases with K18G showing the fastest aggregation followed by K18C > K18 > K18M (Figure 6A). For 2N4R tau, all proteoforms have similar kinetics, except for 359M which is slower (Figure 6B). Using the maximum thioflavin T (Tht) signal, we probed the fibril mass and polymorph distribution for each of the aggregation assays. The intensity of the Tht signal is proportional to the fibril mass, but for different polymorphs, the signal may not be linear for structurally distinct binding sites.⁵⁰ For all K18 proteoforms, the fluorescence signals are of a similar magnitude with the exception of weaker fluorescence for K18M. Tau 359M also has a lower Tht binding than the other tau constructs. These data suggest that the fibrils with comparable Tht signal have a similar mass/polymorph distribution, but the large glycans at N359 have either a different ordered core structure or a less hydrophobic fibril surface due to the large hydrophilic glycan, which prevents Tht binding. To test these hypotheses, we investigated if N-glycosylation causes morphological changes in tau filaments (Figure 6C–F). All the proteoforms generated heterogeneous fibrils, with four different morphology types (straight, jagged, hose, and twisted). We determined the relative proportions of each fibril morphology: the straight fibrils have the highest population among all proteoforms, while hose fibrils are the least common; the jagged fibrils are the next most abundant for K18, while twisted fibrils are the next largest population of tau fibrils. The similarity in fibril distributions points to the notion that 359M has a weaker Tht signal due to steric hindrance of the glycan located on the surface of the fibril. The crossover distances of each fibril morphology type are comparable between K18 and tau. Interestingly, the diameter of the K18 fibrils is inversely correlated with their aggregation kinetics. For 2N4R tau, glycans at N359 form a higher population of structures with smaller diameters than the unmodified or N410 modified tau. From these studies we concluded that N-glycosylation of K18 and 2N4R tau produces heterogeneous fibrils, and each morphology type may have a structurally similar core, based on similarities in their cross over distances.

Figure 6.

Aggregation properties of N-glycosylated K18 and Tau. (A) Normalized heparin-induced aggregation kinetics of K18 glycoproteins and maximum Tht signal during aggregation (50 μM K18, 10 μM heparin, 100 mM NaCl, 10 mM DTT, 25 mM HEPES, 10 μM Tht, pH 7.4, 37 °C; the error bars represent SEM). (B) Normalized heparin-induced aggregation kinetics of 2N4R Tau glycoproteins and maximum Tht signal during aggregation (20 μM Tau, 5 μM heparin, 50 mM NaCl, 5 mM DTT, 25 mM HEPES, 10 μM Tht, pH 7.4, 37 °C; the error bars represent SEM). (C) Negative stain TEM images of the representative fibril morphologies for K18 glycoproteins. The scale bars correspond to 100 nm. (D) Fibril morphology populations generated during the aggregation of glycosylated K18 and Tau. Three random images are selected from three repeats, and the fibril morphology populations are averaged from nine images total. (E) Crossover distance for glycosylated K18 and Tau fibrils. Three random images from three repeats were selected, and the cross over distances were measured along one fibril of each morphology with ImageJ (n = 3, the error bars represent SD). (F) Average diameter of the aggregated K18 samples. Three random images from three repeats were selected and analyzed by Fibril J (the error bars represent SD). (G) BLI of Tau-heparin affinities measured with heparin immobilized on streptavidin (SAX) biosensors and exposed to increasing concentrations of glycoproteins (all experiments were performed in triplicate and the error bars represent SD).

The variations in aggregation kinetics can be attributed to distinct tau-heparin binding affinities, as these interactions are driven by electrostatic forces between positively charged regions in tau (MTBR, PRR, and CTD) and the negatively charged heparin.^51,52 The affinity data measured with biolayer interferometry (BLI) shows that K18 is a weaker binder than the full length 2N4R tau, which points to its reduced number of lysine/arginine residues that contribute to binding (Figure 6G). These results agree with the kinetic data because K18 with small glycans has stronger binding and more rapid aggregation kinetics than K18 with a large glycan or no glycan. Furthermore, the glycosylated tau constructs have similar binding and aggregation kinetics, while the nonglycosylated form has slightly stronger K_D and enhanced aggregation kinetics, supporting the notion that higher tau-heparin interactions are responsible for tau aggregation.

Liquid–liquid phase separation of N-glycosylated tau may concentrate the protein to facilitate oligomerization or fibril nucleation, raising questions about whether glycans increase the propensity to phase separate. Complex coacervation is driven by attractive electrostatic interactions between positively charged regions of tau (MTBR and PRR) and polyanions such as RNA.^53–57 Using a turbidity phase diagram assay, we determined that N-glycosylation regulates polyanion-induced LLPS (Figure 7). Upon mixing with poly(U) RNA, all K18 proteoforms immediately became turbid. Therefore, we constructed a phase diagram at a fixed concentration of K18 but varied the amounts of poly(U) RNA (Figure 7A). The diagrams show that small glycans promote an expanded equilibrium window because K18G and K18C undergo complex coacervation at concentrations of RNA higher than K18. Furthermore, the effect diminishes with increasing sugar size because K18M droplets form at RNA concentrations lower than K18C or K18G. However, N-glycosylation has a net positive effect because K18 complex coacervation peaks and tappers off at lower RNA levels. To probe why small glycans have an expanded phase diagram although they bear the same number of positively charged residues, we used BLI to measure tau-RNA binding (Figure 7L). To ensure all of the binding sites are available, we used 5′-biotin-labeled U(40) immobilized on a streptavidin biosensor probe. The K18-RNA binding is weaker for the constructs with GlcNAc, consistent with its phase diagram because higher molar equivalents of RNA are required to reach charge matching conditions.

Figure 7.

LLPS of N-glycosylated K18 and Tau. (A) Turbidity phase diagram for K18–poly(U) LLPS (30 μM K18, variable μg/mL poly(U), 10 mM DTT, 25 mM HEPES, pH 7.4, 23 °C). (B) Salt resistance assay of K18–poly(U) LLPS (30 μM K18, 60 μg/mL poly(U), variable mM NaCl, 10 mM DTT, 25 mM HEPES, pH 7.4, 23 °C). (C) FRAP kinetics of K18–poly(U) LLPS after 1 h (30 μM K18, 3 μM AF488 labeled K18 (10% labeling), 60 μg/mL poly(U), 5 mM DTT, 25 mM HEPES, pH 7.4, 23 °C). (D) Average droplet area vs time for K18–poly(U) LLPS (30 μM K18, 3 μM AF488-labeled K18 (10% labeling), 60 μg/mL poly(U), 10 mM DTT, 25 mM HEPES, pH 7.4, 23 °C). (E) Determination of C_sat for Tau self-coacervation. Turbidity was measured after 5 min at 400 nm (variable μM Tau, 10% (w/v) PEG-10, 5 mM DTT, 25 mM HEPES, pH 7.4, 23 °C). (F) Salt resistance assay of Tau self-coacervation (15 μM Tau, 10% (w/v) PEG-10, 5 mM DTT, 25 mM HEPES, pH 7.4, 23 °C). (G) FRAP of Tau self-coacervation (15 μM Tau, 0.15 μM AF488 labeled Tau (10% labeling), 10% (w/v) PEG-10, 5 mM DTT, 25 mM HEPES, pH 7.4, 23 °C). (H) Average droplet area vs time for Tau-PEG LLPS. (I) Liquid phase separated droplets of the K18 (after 3 h) and the Tau (after 4 h) constructs (in all images scale bars are 6 μm). (J) Representative images of droplet growth of Tau. (K) Representative images of FRAP recovery of Tau. (L) Tau-U(40) BLI affinity data and the summary for U(40) immobilized on streptavidin (SAX) biosensors and exposed to increasing concentrations of glycoproteins.

Next, we aimed to determine if N-glycosylation impacts cofactor-free LLPS by measuring the saturation concentration (C_sat). Self-coacervation of 2N4R tau occurs in the presence of molecular crowding agents because it is driven by electrostatic interactions between the cationic MTBD and PRR with the acidic N- and C-termini.^58–61 We found that upon mixing of PEG with 2N4R tau (but not K18), a turbid solution formed immediately. Some PTMs such as phosphorylation and FTDP-17 mutations can change the minimal concentration of self-coacervation,^62–64 but N-glycosylation has no effect on C_sat because all proteins underwent LLPS at 2–3 μM in the presence of 10% PEG (Figure 7E). Using a salt resistance assay, we also addressed the question if N-glycosylation impacts the attractive interactions during LLPS (Figure 7B,F). Electrostatically driven phase separation can be weakened by salts that reduce the charge–charge matching interactions between oppositely charged residues.^{53,55,56,60,65} For the complex coacervation of K18, we observed that N-linked glycans promote increased salt resistance because the turbidity is diminished less as compared to the nonglycosylated form. However, in the self-coacervation of 2N4R tau, all the proteoforms tested have similar salt resistance, regardless of the glycosylation state.

To assess the mobility of tau droplets we used fluorescence recovery after photobleaching (FRAP). Bleaching of tau droplets was rapid, and the recovery occurred over several minutes, indicating that the droplets have a liquid-like character (Figure 7C,G). During complex coacervation, single GlcNAc promotes gelation or Maxwell glassification,⁶⁶ but the effect diminishes with larger sugars in K18G and K18C having larger immobile fractions as compared to K18 and K18M. During self-coacervation of 2N4R tau, we also observed a similar phenomenon at N410 as GlcNAc formed less mobile droplets, while the high-mannose structure produced droplets with higher mobility than unmodified tau. Glycosylation at N359 does not change the dynamics or mobility of self-coacervated tau. To investigate the morphology of tau droplets, time dependent microscopy experiments were performed (Figure 7I–K). For complex coacervation, all liquid droplets were round, regardless of the proteoform, the droplet area increased and plateaued after 3 h to a similar extent. For self-coacervation, we observed that the glycans at N410 impact the droplet morphology with large sugars favoring less dense droplets and small glycans having the opposite effect.

DISCUSSION

The primary aim of this study was to elucidate the role of N-glycosylation in the pathogenicity of tau protein. Employing a molecular approach, we conducted a systematic investigation to determine whether the position and size of the glycan moieties influence the biophysical attributes of tau. This research provides insights into the mechanisms of tau propagation and its impact on neuronal function.⁶⁷ The high mannose glycan, observed in AD patients with high frequency, was chosen for this study due to its biological prevalence. Additionally, single GlcNAc residues were utilized to explore the effects of N-glycosylation, acting as a model to capture the local interactions, similar to K18 used as a surrogate of full-length tau. These glycans were synthesized chemically, with oligosaccharide amines incorporated onto peptides via the Lansbury aspartylation—a method that enables site-specific glycan addition, crucial for addressing the questions posed by this study. Aspartylation at N359 is the most challenging step in the synthesis, as the ligation does not scale up well, and multiple 1 mg scale reactions were required. However, the same process at N410 proceeded efficiently even on scale. To achieve high purity of the material, we optimized the synthesis with ~30-mer peptide fragments, which present fewer challenges as compared to longer peptide segments. The accumulation of minor impurities from SPPS that have identical retention times as the product mass can be difficult to manage, and great care was taken to remove them. Furthermore, the impurities can carry forward and become multiplied after each ligation, resulting in poorly resolved mass signals and impurities inseparable with the conventional reverse-phase HPLC. We overcame these issues using enabling purification technology, sample displacement chromatography. In this method, the analyte competes for binding sites on the stationary phase, and the impurities with subtle differences in polarity and size can be separated. Sample displacement chromatography provides a significant advantage in purifying peptides and proteins due to its unique mechanism of sequential adsorption based on differential affinities allowing stronger-binding target proteins or peptides to displace weaker-binding impurities under overloading conditions. This method is particularly advantageous when dealing with complex mixtures, as it enables effective fractionation even under high loading conditions. Our separations were run under saturation where a C18 column was loaded with 5–20 mg of the peptide and eluted with a 5–45% MeCN gradient over 60 min. The elution was run over 10–15 min, and 0.5–1.0 mL fractions were collected with ~30% of the pure fractions. Once purified, the lyophilized peptides and protein fragments were stored under standard conditions (−20 or −80 °C and protected from light).

In solution, all tau proteins remained largely disordered, and N-glycosylation was observed to alter protein mobility on gel electrophoresis. Our results indicate that N-glycosylation of the 2N4R tau isoform does not significantly alter its ability to promote microtubule polymerization. However, the presence of high mannose glycans at N410 was found to facilitate the phosphorylation of adjacent serine residues by GSK3β, a kinase implicated in the hyperphosphorylation of tau, thereby underscoring the potential pathological implications of N410 glycosylation in tau-related diseases.

The properties of the bulk material are also altered by N-glycosylation. Glycans in PHFs may cause their characteristic twisted morphology because deglycosylation of the fibrils converts them into bundles of straight filaments.⁸ However, our glycosylated fibrils do not exhibit a helical appearance.^68–74 In postmortem samples, the insoluble fraction of tau contains more truncated and fucosylated glycans than the soluble fraction which has larger high mannose glycans.^8,9 This result is consistent with our observation that small N-linked glycans at N359 promote aggregation, whereas large glycans at N359 disfavor fibrillation. However, these kinetic differences may not be sufficient to rationalize why differences in glycan composition between soluble and aggregated forms exist because modifications at the N359 position are in the ordered AD core and would interfere with the assembly of PHFs.

Tau does not undergo aggregation under normal conditions because it is a cationic, disordered, and highly soluble protein. However, charge neutralization of tau induces fibrilization, either by hyperphosphorylation or by polyanionic cofactors such as heparin and RNA.^5,75–80 Our studies employ heparin to induce tau aggregation, although it must be noted that heparin becomes incorporated into the in vitro fibril, and the cryo-EM structure of polyanion-induced aggregates may not resemble disease-specific folds.^69,70,72,74

Glycans can mediate the phase separation of tau. Two charge–charge matching mechanisms of LLPS of tau are known either through tau-tau (self-coacervation) or tau-polyanion (complex coacervation) interactions.^56,66 Hydrophobically driven self-coacervation of tau can also occur upon phosphorylation or exposure to high salt concentrations.^59,81,82 Previous studies have shown that hyperphosphorylation and pseudophosphorylation of tau promote its self-coacervation but diminish its electrostatic complex coacervation.^54,59,61,63 It remained unknown how N-glycosylation, a neutral post-translational modification, regulates tau phase separation. Our findings indicate that small glycans at N359 and N410 reduce tau mobility within liquid droplets, with large glycans reversing this effect. Complex coacervation of tau is also a function of temperature supporting the earlier notion that LLPS is entropy-controlled and driven by release of water. Glycans as polyhydroxy compounds are excellent binders of water, thus dehydration is less favorable for glycoproteins. N-linked glycans also promote complex coacervation because they enable LLPS at higher concentrations of RNA than the nonglycosylated form, and this effect diminishes with increasing sugar size. Moreover, we show that N-glycosylation promoted an increased salt stability during complex coacervation but had no effect on self-coacervation, suggesting that N-glycosylation helps maintain the pairing with RNA. In a broader sense, the marked changes in the biophysical properties may translate into the increased uptake of N410 aggregated tau proteins into human neurons, which may expose unique residues of tau to engage with the binding receptors. These findings suggest a model where tau glycosylation leads to increased phosphorylation, enhanced tau aggregation, and amyloid formation (Figure 8). Extracellularly, glycosylated tau exhibits increased stability and facilitates its spread across neuronal networks, contributing to disease progression. Understanding glycan-mediated mechanisms not only addresses fundamental biological questions but also informs the development of targeted therapies, including immunotherapies aimed at modulating tau glycosylation and other PTMs.⁸³

Figure 8.

Schematic illustration of the proposed mechanism of Tau glycosylation and the impact of glycan attachment on biophysical properties and cellular responses. Upward arrows signify increased properties, downward arrows denote decreased propensity, and horizontal arrows represent no significant change from unmodified 2N4R Tau.

CONCLUSIONS

This study describes a comprehensive analysis of the role of N-glycosylation in tau and potential role in AD pathogenesis. The unique effects of glycan attachment at N410 provides strong evidence that perturbation of glycan processing may lead to increased amyloid formation and facilitate the spread of the pathology.

ABBREVIATIONS

Acm

Acetamidomethyl

AcOH

Acetic acid

Ac₂O

Acetic anhydride

AgOAc

Silver(I) acetate

AgOTf

Silver trifluoromethanesulfonate

BH₃·THF

Borane-tetrahydrofuran complex

BLI

Biolayer interferometry

Bu₂BOTf

Di-tert-butyl tifluoro(4,4,5,5-tetramethyl-1,3,2-dioxaborolane-2-oxide)borate

CD

Circular Dichroism

C_sat

Saturation concentration

DIPEA

N,N-Diisopropylethylamine

DMF

Dimethylformamide

DPDS

Diphenyldiselenide

DSL

Diselenide-selenoester-ligation

DTT

Dithiothreitol

EGTA

Ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N’-tetraacetic acid

ESI

Electrospray Ionization

FRAP

Fluorescence Recovery After Photobleaching

Gnd·HCl

Guanidine hydrochloride

GSH

Glutathione

GTP

Guanosine triphosphate

HATU

Hexafluorophosphate azabenzotriazole tetramethyl uronium

HEPES

4-(2-Hydroxyethyl)-1-piperazineethanesulfonic acid

HRMS

High Resolution Mass Spectrometry

LLPS

Liquid–liquid phase separation

MeOH

Methanol

MeONH₂·HCl

Methylamine hydrochloride

MESNa

Sodium 2-mercaptoethanesulfonate

MPAA

Mercaptopropionic acid

NaOMe

Sodium methoxide

NCL

Native Chemical Ligation

NIS

N-Iodosuccinimide

PEG-10

Polyethylene glycol 10

PHF

Paired Helical Filament

PIPES

Piperazine-N,N’-bis(2-ethanesulfonic acid)

PMFS

Phenylmethylsulfonyl fluoride

SAX

Streptavidin

SD

Standard deviation Sodium Dodecyl Sulfate-Polyacrylamide Gel

SDS-PAGE

Electrophoresis

SEM

Standard error of the mean

TCEP·HCl

Tris(2-carboxyethyl)phosphine hydrochloride

TEM

Transmission Electron Microscopy

TFA

Trifluoroacetic acid

Thz

Thiazolidine

Tht

Thioflavin T

TIPSH

Triisopropylsilane